Discussing the limits of artificial intelligence


It’s hard to visit a tech site these days without seeing a headline about deep learning for X, and that AI is on the verge of solving all our problems. Gary Marcus remains skeptical.

Marcus, a best-selling author, entrepreneur, and professor of psychology at NYU, has spent decades studying how children learn and believes that throwing more data at problems won’t necessarily lead to progress in areas such as understanding language, not to speak of getting us to AGI – artificial general intelligence.

Marcus is the voice of anti-hype at a time when AI is all the hype, and in 2015 he translated his thinking into a startup, Geometric Intelligence, which uses insights from cognitive psychology to build better performing, less data-hungry machine learning systems. The team was acquired by Uber in December to run Uber’s AI labs, where his cofounder Zoubin Ghahramani has now been appointed chief scientist. So what did the tech giant see that was so important?

In an interview for Flux, I sat down with Marcus, who discussed why deep learning is the hammer that’s making all problems look like a nail and why his alternative sparse data approach is so valuable.

We also got into the challenges of being an AI startup competing with the resources of Google, how corporates aren’t focused on what society actually needs from AI, his proposal to revamp the outdated Turing test with a multi-disciplinary AI triathlon, and why programming a robot to understand “harm” is so difficult.

Subscribe to our Channel